Comparison between statistical and principal component analysis in reduction of near-field FDTD data

Abstract

We present a comparison of statistical and principal component analyses (PCA) for reducing the amount of near-field data required as an input to a time-domain spherical-multipole near-to-far-field transformation. Such transformations are necessary for finite-difference time-domain (FDTD) simulations that typically only model the near-field. The authors demonstrate their approach for the case of far-fields scattered by a dielectric sphere. For a threshold value of 106 the PCA technique reduced the data required by 32%, using 12 components. A similar compression was achieved with statistical analysis, when the threshold is set at 10-7. Their work shows that the proposed statistical compression is preferable because it has simpler implementation and low-cost processing, when implemented together with the NFF transformation code.